fix(usage): read Gemini usageMetadata out of the antigravity response envelope (#10430)

* fix(usage): read Gemini usageMetadata out of the antigravity response envelope

Port decolua/9router#59d858b: antigravity/gemini-cli wrap non-streaming
payloads in { response: {...} }, so extractUsageFromResponse only saw the
top-level usageMetadata and every non-streaming antigravity request logged
zero usage (IN 0 | OUT 0) and zeroed usage-dashboard rows. Top-level
metadata keeps priority; OpenAI/Claude branches untouched.

* chore(changelog): fragment for #10430 antigravity usage envelope

* ci: re-run dast-smoke (Build CLI bundle runner timeout flake)

---------

Co-authored-by: Rouzbeh <rqzbeh@users.noreply.github.com>

changelog.d/fixes/10430-antigravity-usage-envelope.md

@@ -0,0 +1 @@
+- **fix(usage):** read Gemini `usageMetadata` out of the antigravity `{ response: {...} }` envelope so non-streaming requests log real token usage instead of `IN 0 | OUT 0` (port of decolua/9router#59d858b) ([#10430](https://github.com/diegosouzapw/OmniRoute/pull/10430)) — thanks @rqzbeh

open-sse/handlers/usageExtractor.ts

@@ -93,14 +93,17 @@ export function extractUsageFromResponse(responseBody, provider) {
     };
   }
 
-  // Gemini format
-  if (responseBody.usageMetadata && typeof responseBody.usageMetadata === "object") {
+  // Gemini format. Antigravity / gemini-cli wrap the payload in
+  // { response: { ... } } — read the envelope so non-streaming requests do
+  // not silently log zero usage (port of decolua/9router#59d858b).
+  const usageMetadata = responseBody.usageMetadata || responseBody.response?.usageMetadata;
+  if (usageMetadata && typeof usageMetadata === "object") {
     // Gemini reports thoughts outside candidates. Fold them into completion so
     // every provider keeps reasoning as a subset of completion tokens.
-    const thoughts = responseBody.usageMetadata.thoughtsTokenCount || 0;
+    const thoughts = usageMetadata.thoughtsTokenCount || 0;
     return {
-      prompt_tokens: responseBody.usageMetadata.promptTokenCount || 0,
-      completion_tokens: (responseBody.usageMetadata.candidatesTokenCount || 0) + thoughts,
+      prompt_tokens: usageMetadata.promptTokenCount || 0,
+      completion_tokens: (usageMetadata.candidatesTokenCount || 0) + thoughts,
       reasoning_tokens: thoughts,
     };
   }

tests/unit/usage-extractor.test.ts

@@ -232,6 +232,46 @@ test("extractUsageFromResponse reads Gemini usageMetadata and thinking tokens",
   });
 });
 
+test("extractUsageFromResponse reads Gemini usageMetadata from the antigravity response envelope", () => {
+  // Antigravity / gemini-cli wrap non-streaming payloads in { response: {...} }
+  // (port of decolua/9router#59d858b — previously logged zero usage).
+  const usage = extractUsageFromResponse(
+    {
+      response: {
+        usageMetadata: {
+          promptTokenCount: 42,
+          candidatesTokenCount: 13,
+          thoughtsTokenCount: 4,
+          cachedContentTokenCount: 7,
+        },
+      },
+    },
+    "antigravity"
+  );
+
+  assert.deepEqual(usage, {
+    prompt_tokens: 42,
+    completion_tokens: 17,
+    reasoning_tokens: 4,
+  });
+});
+
+test("extractUsageFromResponse prefers top-level usageMetadata over the envelope", () => {
+  const usage = extractUsageFromResponse(
+    {
+      usageMetadata: { promptTokenCount: 1, candidatesTokenCount: 2 },
+      response: { usageMetadata: { promptTokenCount: 99, candidatesTokenCount: 99 } },
+    },
+    "gemini"
+  );
+
+  assert.deepEqual(usage, {
+    prompt_tokens: 1,
+    completion_tokens: 2,
+    reasoning_tokens: 0,
+  });
+});
+
 test("extractUsageFromResponse returns null when usage is missing", () => {
   const usage = extractUsageFromResponse(
     {